A general strategy for reducing the variability in response-adaptive designs

Adaptive designs for clinical trials address the ethical requirement of allocating fewer patients to the inferior treatment as the trial progresses. We present here a generic strategy partly motivated by play-the-winner heuristic for reducing the variability of such designs. It is shown through simulations from a real clinical setting that employing this strategy improves the safety and the performance of some standard response adaptive designs. A closely related alternate strategy amenable to a theoretical analysis is also proposed with the same objective of lower variability.